Unveiling a novel transient druggable pocket in BACE-1 through molecular simulations: Conformational analysis and binding mode of multisite inhibitors

Abstract

The critical role of BACE-1 in the formation of neurotoxic ß-amyloid peptides in the brain makes it an attractive target for an efficacious treatment of Alzheimer's disease. However, the development of clinically useful BACE-1 inhibitors has proven to be extremely challenging. In this study we examine the binding mode of a novel potent inhibitor (compound 1, with IC50 80 nM) designed by synergistic combination of two fragments-huprine and rhein-that individually are endowed with very low activity against BACE-1. Examination of crystal structures reveals no appropriate binding site large enough to accommodate 1. Therefore we have examined the conformational flexibility of BACE-1 through extended molecular dynamics simulations, paying attention to the highly flexible region shaped by loops 8-14, 154-169 and 307-318. The analysis of the protein dynamics, together with studies of pocket druggability, has allowed us to detect the transient formation of a secondary binding site, which contains Arg307 as a key residue for the interaction with small molecules, at the edge of the catalytic cleft. The formation of this druggable "floppy" pocket would enable the binding of multisite inhibitors targeting both catalytic and secondary sites. Molecular dynamics simulations of BACE-1 bound to huprine-rhein hybrid compounds support the feasibility of this hypothesis. The results provide a basis to explain the high inhibitory potency of the two enantiomeric forms of 1, together with the large dependence on the length of the oligomethylenic linker. Furthermore, the multisite hypothesis has allowed us to rationalize the inhibitory potency of a series of tacrine-chromene hybrid compounds, specifically regarding the apparent lack of sensitivity of the inhibition constant to the chemical modifications introduced in the chromene unit. Overall, these findings pave the way for the exploration of novel functionalities in the design of optimized BACE-1 multisite inhibitors.

Fig 1. Chemical structures of the huprine-rhein hybrids 1 and 2 and of the 6-chlorotacrine-chromene hybrid 3.

Fig 2. Representation of the multistep computational approach utilized in the focusing search strategy followed in this study.

Fig 3. Representation of BACE-1 as found in the structures 1SGZ and 2P4J.

The backbone of the superposed proteins is shown as white cartoon, and the ligand located along the catalytic cleft in 2P4J is shown as cyan-colored sticks. The flap region (blue) shows the open and closed conformations typically found in apo and substrate-bound states, respectively. The other flexible region is located in a region distal from the catalytic site and is shaped by residues 8–14 (red), 154–169 (green), and 307–318 (magenta). The two major conformations found for the loop formed by residues 8–14 is shown. Loops 154–169 and 307–318 are generally disordered in other PDB structures.

Fig 4. RMSD profile for the ensemble of 100 trajectories started from X-ray structures (A) 1SGZ and (B) 2P4J.

The dark blue line reflects the average value of the RMSD (Å) determined for the C-alpha atoms of the whole protein for the ensemble of snapshots collected at a given time (ns) relative to the energy-minimized X-ray structures (the light blue line denotes the variance of the averaged RMSD). The green line reflects the RMSD (Å) determined for the C-alpha atoms of the average structure obtained from the snapshots sampled at a given time (ns) relative to the X-ray structure.

Fig 5. Principal Component Analysis of the MD trajectories.

(A) Projection of the ensemble of snapshots collected along the last 50 ns of the 200 MD trajectories on the space defined by the two major principal components (PC1, PC2). (B, C) Representation of the major deformations in loops 8–14, 154–169, and 307–318 generated by the two principal eigenvectors. Deformations are projected onto the average structure of the protein backbone, which is shown as white cartoon with the only exception of the flap, which is shown in blue.

Fig 6. Representative snapshots of BACE-1 with the loop 8–14 in the out conformational state.

Structures were taken from clusters 13 (1.5%; A), 22 (0.8%; B), 25 (0.5%; C), and 26 (0.4%; D). The flap region is shown in blue, and the loops defined by residues 8–14, 154–169, and 307–318 are shown in red, green and magenta, respectively. The size of the secondary pocket in the selected snapshots is around (A) 460, (B) 520, (C) 770, and (D) 1,250 ų.

Fig 7. Representation of the druggable sites identified from fPocket analysis.

The BACE-1 backbone is shown as white cartoon. (A) The catalytic and secondary pockets found in cluster 13 are shown as orange and yellow spheres. (B) The single pocket found along the catalytic cleft in cluster 21 is shown as yellow spheres. The flap region is shown in blue, and the loops defined by residues 8–14, 154–169, and 307–318 are shown in red, green and magenta, respectively. The ligand located along the catalytic cleft in structure 2P4J is shown as cyan-colored sticks.

Fig 8. Surface representation of the best pose obtained for the docking of rhein to the secondary pocket in clusters (A) 13 and (B) 21.

The BACE-1 backbone is shown as white cartoon. The flap region is shown in blue, and the loops defined by residues 8–14, 154–169, and 307–318 are shown in red, green, and magenta, respectively. Arg307 is shown as sticks. The inset depicts the location of the binding pocket in the whole protein.

Fig 9. Superposition of the snapshots collected at the end of the trajectories run for the BACE-1 complexes with hybrid compounds (A) (+)-1, (B) (–)-1, and (C) 3.

The protein backbone is shown as white cartoon. The flap region is shown in blue, and the loops defined by residues 8–14, 154–169, and 307–318 are shown in red, green and magenta, respectively. Arg307 is shown in sticks.